Objective. Under the background of two-way referral in medical consortium, the provision of high-quality hierarchical care requires active integration and cross- organizational sharing of patient data. However, cross-organizational data sharing among hospitals is not always spontaneous, comprehensive, and effective.
Methods. This paper reports on a research study aiming to identify the barriers of patient data sharing. Due to the exploratory nature of this study, two medical consortiums in central China were chosen to conduct a case study. Specifically, a mixed qualitative-quantitative approach was employed for the case study analysis and verification.
Conclusions. The analysis of qualitative interview data pointed to 27 barriers of data sharing under three main themes: psychological ownership; territorial behaviours; cross-organizational dynamism. The quantitative Kano analysis revealed the manifestation of these barriers, and their prioritization were classified into four types: must-be, one-dimensional, attractive, and indifference.
Implications. This study provides valuable references for enhancing the effectiveness of patient data sharing and optimizing the allocation of healthcare resources. The case setting is based on China’s medical consortium, but the findings offer useful insights and indications that can be shared across international borders.
Introduction
Data, commonly understood as statistical observations or collected evidence (Zins, 2007), has become a crucial asset in modern societies (Hilbert, 2016). However, data is often only used to a limited extent or frequently only in an internal organization context (Prajogo et al., 2018). Data sharing refers to breaking down the data barriers between organizational departments, establishing a unified data sharing mechanism, and accelerating the flow of data resources within the organization (Harris et al., 2007). It also encompasses making research data publicly available with minimal reuse restrictions, known as open data (Thoegersen & Borlund, 2022). While data sharing increases the complexity and quantity of data (Kowalczyk & Shankar, 2011), it plays a vital role in advancing scientific collaboration (Priego et al., 2022), urban governance (Palm et al., 2019), public health management (Brand et al., 2022), and social development.
The hospitals are typically data-intensive organizations. Health decision making has become increasingly complex, and the use of patient data has become essential in this information age (Van, 2014). Integration of patient data has been perceived to be a breakthrough for the discovery of timely, cost-efficiency therapies, and a crucial step towards enhancing the quality of medical care for curing diseases (Weitzman et al., 2010). Two-way referral serves as a classic example of patient data sharing. It refers to the transfer process between hospitals of different levels, specialized hospitals, or between general and specialized hospitals (Huang et al., 2023) based on the health needs of the population. Nowadays, the integration and cross-organizational sharing of patient data has become a key part of achieving the goals of hierarchical diagnosis and two-way referral.
Recognizing the importance of addressing these medical needs and challenges, the Chinese government attempted to improve medical development through the establishment of medical consortium (Li et al., 2017). Medical consortium includes urban medical groups led by tertiary public hospitals, county medical communities led by county hospitals, and specialist alliances and telemedicine collaboration networks with specialist associations as the link (Cai et al., 2018).
Nevertheless, barriers exist to integrated working, including technical and ethical issues, organizational resistance, financial costs, and privacy concerns (Houtkoop et al., 2018; Li & Cong, 2024; Reddy et al., 2024). Meanwhile, different hospitals have different regulations and willingness on sharing patient data in the two-way referral process, which has made the implementation of hierarchical diagnosis and treatment less effective than expected. Yet current research in this field is notably limited.
The paper reports on a research project that aimed to identify and solve existing barriers to cross- organizational sharing of patient data in the context of two-way referral in medical consortium. To further the analysis, the following research questions were formulated:
RQ1. What are the barriers of patient data sharing?
RQ2. Which are the priorities of these barriers?
RQ3. What are the practical strategies to address these barriers?
Research methodology and design
To answer our research questions, and considering the exploratory nature of this study, a case study approach is employed. According to China’s current ‘Hospital Classification and Management Measures’, 3A hospitals are the highest level in the classification of hospitals in Mainland China. After careful consideration, Xiangyang First People’s Hospital Medical Consortium and Xiangyang Traditional Chinese Medicine Hospital Medical Consortium is adopted as the research case. Both hospitals are Grade 3A hospitals, which ensures the representativeness of the research and can accurately reflect the situation in the entire medical field.
The research was carried out in two stages. In stage one, qualitative research was undertaken to identify the barriers of patient data sharing. Semi-structured interviews were used to gather qualitative data, with guaranteed access to interview respondents. All interviews were recorded using a digital recorder and transcribed verbatim into MS Word files. NVivo 14 was used to facilitate the analysis of interview data.
After that, a Kano model analysis was carried out in stage two to verify and validate the barriers identified in the qualitative study. In this stage, Kano model analysis was used to prioritize barriers of patient data sharing. It was expected that through the prioritization processed, effective strategic suggestions could be formulated.
Stage one: qualitative researchResearch methods and processes
As discussed previously, semi-structured interviews were used to collect qualitative data, which were then analysed using a thematic analysis approach. A framework proposed by Brown et al. (2014) was used as a basis for data collection and analysis, as shown in Figure 1. The framework comprises three categories which are psychological ownership, territorial behaviours, and cross- organizational dynamism (Brown et al., 2014), that can help understanding barriers in patient data sharing.
Psychological ownership: hospitals have invested a lot of manpower, material resources and time costs to maintain patient data, so that they have regarded patient data as their inseparable assets.
Territorial behaviours: are actions can be adopted by hospitals for the protection of their perceived ownership over data. Hospitals are inclined to keep and safeguard their data instead of sharing it willingly.
Cross-organizational dynamism: can moderate the influence of psychological ownership on territorial behaviours.
A preliminary analytical model based on the research of Brown et al. (2014)
none
The semi-structured question script was designed based on Figure 1 and consisted of 15 questions, including 1 ice-breaking question and 14 theoretical probing questions. Before entering the case study, the question script was proofread and trialled with two professors in the field of cross- organizational data sharing. A total of 13 staff and patients were approached and interviewed. The demographic profile of the participants is illustrated in the Table 1 below:
Information on interview participants
No.
Interviewer
Department/Section
A
Engineer
Information Center
B
Engineer
Information Center
C
Department Leader
Information Center
D
Clerk
Medical Office
E
Clerk
Medical Office
F
Department Leader
Information Center
G
Engineer
Information Center
H
Department Leader
Medical Office
I
Head Nurse
Nephrology
J
Nurse
Nephrology
K
Chief Physician
Nephrology
L
Referral patient
Orthopedics
M
Referral patient
Neurosurgery
All interviews were interviewed in March 2024. At the beginning of every interview, the participant was consulted if the use of digital recorder could be accepted. The participants were assured that they can stop the recorder whenever they feel uncomfortable.
Instead of directly diving into the question-answer dialogue, the interviews started by making some irrelevant and light conversations. After obtaining informed consent, all interview recordings were then transcribed verbatim by the research team. The transcripts were analysed using a thematic analysis approach.
In the research, the Brown et al. (2014)’s framework was considered as generic requirements and were converted into a preliminary set of codes in the early stages. Throughout the data analysis, these codes were reconceptualized, expanded and revised, according to the statements, interpretations, and perspectives of the individual interview participants.
Findings
The analysis of the interview data inherited the thematic analysis strategy of the model (Figure 1), the themes and codes resulting from the analysis are shown in the Table 2.
Requirements, themes, coding, and identifier
Theme
Sub-theme
Coding
Identifier
Psychological Ownership
-
Fear of declining status after sharing
A1
Reluctance to give up data assets
A2
Lack of access to required data
A3
Territorial Behaviours
Identity-Oriented Marking
Using different data systems
B1
Different data processing standards
B2
Control-Oriented Marking
Complex approval process for declarations
C1
Self-determination of the form of sharing
C2
Anticipatory Defences
Distrust of shared patient data
D1
Passive response to data sharing construction
D2
Insufficient resources in primary medical institutions
D3
Lack of personnel in data sharing positions
D4
Reactionary Defences
Excessive accountability for data accidents
E1
Cross-Organizational Dynamism
Cross-Organizational Management
Deviation between government planning and real needs
F1
Lack of clear policies and regulations
F2
Lack of government supervision and guidance
F3
Limited results of medical insurance reform
F4
Cross-Organizational Relationships
Lack of high-level cross-organizational mechanism
G1
Lack of preliminary cooperation foundation
G2
Lack of trust between organizations
G3
Interest game of tertiary hospitals
G4
Loose cooperation of medical consortium
G5
Cross-Organizational Risk
Unclear rights and responsibilities for sharing
H1
Lack of incentive and compensation mechanism after sharing
H2
Low level of informatization of medical institutions
H3
Lack of financial support
H4
Concerns about data quality
H5
Concerns about data security
H6
The data analysis shows that the interviewees frequently highlighted psychological barriers related to the perceived ownership of data assets. A significant challenge for hospitals is the fear of diminished status after data sharing. This concern is particularly pronounced among community and secondary hospitals, as an interviewee noted, ‘liberalising data sharing will enable doctors at the next level of care to improve their standards, which poses a threat to us’ (Interview F). Besides, in the context of big data, ‘data are considered assets’ (Interview A). Hospitals invest considerable resources, including manpower, material, and time, into maintaining patient data, treating it as a crucial and inseparable asset. As a result, they are unwilling to engage with or learn how to use data-sharing platforms, which further complicates their ability to access data through these platforms.
The data analysis showed that territorial behaviours can significantly affect patient data sharing, with four specific manifestations identified. Identity-oriented marking refers to the fact that ‘each
hospital has its own HIS system’ (Interviewee B), enhancing the complication of integrating systems to achieve data sharing. Besides, ‘data processing standards’ (Interviewee A) varies, with no unified stipulation to realize data cross-authentication. Control-oriented marking is evident in the lengthy approval processes of data sharing request, with ‘several people need to sign only a piece of file and taking at least three working days to approve it’ (Interview M). Anticipatory defences especially reflect a passive response and prudent attitude to obtained data. Meanwhile, deliberate procrastination can also be a common action to defend settled data sharing request, since ‘they even do not arrange people to specifically designated to handle this issue’ (Interview E). Reactionary defences generally occur when hospitals actually experience data infringement by others. There is evidence showing that an excessive accountability for data accidents will be taken as a defensive action to the specific hospital. As a doctor stated that ‘if they fail to take the necessary security measures and cause a data breach, they must take the responsibility’ (Interview L).
Furthermore, the interviews revealed significant barriers to cross-organizational data sharing, primarily related to management, relationships, and risk. The most important barrier for management issue is the current legal framework lacks clarity, as noted by an interviewee, ‘the overall plan does not conform to the actual situation’ (Interview G). To enable effective data sharing management, a unified data centre is essential. However, the absence of a high-level cross- organizational mechanism exacerbates this problem. In terms of relationships, hospitals express concerns about ‘data security’ (Interview B/C/J) and ‘data quality’ (Interview H/I/J), which affect the trust and collaboration for data sharing. Additionally, issues following data sharing include ‘unclear sharing rights and responsibilities’ (Interview F) and an ‘absence of an incentive compensation mechanism’ (Interview A). These risks highlight the need for clearer guidelines and effective incentives.
Stage two: Kano model analysisResearch methods and processes
Kano model is composed of two important concepts, one is the objective realization of requirements, the other is the subjective satisfaction of users. Stage two employs the Kano model to validate and verify the barriers identified in stage one. Moreover, stage two measures the priority of the barriers and classifies the barriers into four categories (sorted in the order they should be solved): Must-be (M); One-dimensional (O); Attractive (A); Indifferent (I) (Xu et al., 2009). Data collection is divided into three main steps.
Questionnaire design. This study transforms the identified disorders into demand items. Excerpts from the interview transcripts from Stage One were used as evidence for the design of questionnaire.
Grading of questions. Every option setting refers to the Likert 5-point scale method (Adelson & McCoach, 2010). And respondents can choose from one of the following five indicators: I like it that way, it must be that way, I am neutral, I can live with it that way, and I dislike it that way.
Questionnaire distribution. This study uses both online and offline methods to place and collect questionnaires. The questionnaires were distributed and returned in April 2024, yielding 98 valid responses.
Findings
Defining primary quality attribute (PQA) was the first step in Kano model analysis. In the analysis, responses from the questionnaire survey to the functional and non-functional questions of the requirements were combined and then assigned to a Kano evaluation matrix (Table 3). Using the matrix, a requirement can be defined and categorized into one of the M, O, A, I, R and Q categories. For instance, the analysis of requirements B1 is shown in Table 4. B1 is an A requirement because A represents the highest ratio. Repeating the above analysis processes, all requirements could be calculated.
Kano evaluation matrix
Requirements
Non-functional issues
Scale
I dislike it that way
I can live with it that way
I am neutral
It must be that way
I like it that way
Functional questions
I dislike it that way
Q
R
R
R
R
I can live with it that way
M
I
I
I
R
I am neutral
M
I
I
I
R
It must be that way
M
I
I
I
R
I like it that way
O
A
A
A
Q
A sample of Kano evaluation matrix for B1
Requirements
A
O
M
I
R
Q
B1
48.0%
10.2%
18.4%
23.5%
0
0
User satisfaction coefficient (UCS) measures a situation whereby achieving a requirement of patient data sharing can prevent dissatisfaction. It comprises two indicators: satisfaction index (SI) and dissatisfaction index (DSI).
SI=A+0A+0+M+IDSI=−1×0+MA+0+M+1
Based on the calculation of SI and DSI, a satisfaction impact matrix can be developed with an X- axis presenting the DSI values and a Y-axis illustrating the SI values.
Therefore, the UCS is developed as shown in Figure 2. Moreover, the barriers are divided into four sections: Attractive (A): B1, C1, D4, E1, F2, G2, G4, G5, H4. One-dimensional (O): B2, H2, H6. Must- be (M): A2, A3, D2, F1, G1, H1, H3, H5. Indifferent (I): A1, C2, D1, D3, F3, F4, G3.
User satisfaction impact matrix
noneDiscussions and conclusions
This paper uses a comprehensive approach to identify barriers to patient data sharing and validate the prioritization of barriers. Using deductive qualitative methods, 27 barriers of data sharing were identified under three main themes. Subsequently, these barriers were categorized into four levels using Kano analysis based on the quantitative measurement. This study generates the following specific strategic suggestions.
First, establishing standardized patient data specifications is a critical prerequisite for effective data sharing. It is recommended to systematically categorize data types, encompassing essential attributes such as format, storage methods, and access rights. A comprehensive and compatible standardization framework should be developed. Based on this, a specialized team should be established to oversee the implementation of these data standards. Additionally, hospitals should prioritize the development of new technologies and increase capital investment to ensure the real-time updating and seamless transmission of critical medical information.
Second, to enhance the diagnostic and therapeutic capabilities of medical consortium, it is crucial to prioritize comprehensive training programs to ensure all employees are equipped with the essential skills and knowledge. This training should focus on fostering a robust understanding of collaboration and resource sharing within the network. As this awareness deepens, employees will gain a deeper understanding of data sharing. This understanding will ultimately foster a cooperative culture, enabling smooth and effective sharing of medical data and resources across organizations.
Third, to address the challenges of financial support and benefit distribution in data sharing, it is imperative to enhance government involvement. Specifically, the government should urgently increase financial support and establish a dedicated subsidy system for patient data sharing. Additionally, a cross-organizational communication framework centered on data sharing should be developed, with a committee comprising hospital representatives overseeing this initiative.
Fourth, for the sake of improving the staff’s ability to understand and utilize the medical consortium effectively, hospitals should implement targeted educational activities emphasizing the importance of data quality, including holding lectures, seminars, etc. At the same time, hospitals should collaborate with data experts and software developers to design comprehensive training courses for the medical information system, covering system functions, operating logic, application scenarios, etc.
In conclusion, cross-organizational patient data sharing in medical consortium offers a unique opportunity for hospitals to redefine, reposition and reinvent themselves. This research adds to the general knowledge base by proposing a comprehensive set of barriers of patient data sharing and strategic suggestions, which can be adopted as a knowledge base and a starting point for future studies. Also, future work can aim at expanding the sample coverage and conducting research on different types of medical consortium.
ReferencesAdelsonJ. L.McCoachD. B.2010Measuring the Mathematical Attitudes of Elementary Students: The Effects of a 4-Point or 5-Point Likert-Type ScaleEducational and Psychological Measurement705796807https://doi.org/10.1177/0013164410366694BrandD.SinghJ. A.McKayA. G. N.CengizN.MoodleyK.2022Data sharing governance in sub-Saharan Africa during public health emergencies: Gaps and guidanceSouth African Journal of Science1181112Article #13892https://doi.org/10.17159/sajs.2022/13892BrownG.CrossleyC.RobinsonS. L.2014PSYCHOLOGICAL OWNERSHIP, TERRITORIAL BEHAVIOR, AND BEING PERCEIVED AS A TEAM CONTRIBUTOR: THE CRITICAL ROLE OF TRUST IN THE WORK ENVIRONMENTPersonnel Psychology672463485https://doi.org/10.1111/peps.12048CaiM.LiuE.TaoH. B.QianZ. M.FuQ.LinX. J.NiZ. L.2018Does A Medical Consortium Influence Health Outcomes of Hospitalized Cancer Patients? An Integrated Care Model in Shanxi, ChinaInternational Journal of Integrated Care182Article 7https://doi.org/10.5334/ijic.3588HarrisD.KhanL.PaulR.ThuraisinghamB.2007Standards for secure data sharing across organizationsComputer Standards & Interfaces2918696https://doi.org/10.1016/j.csi.2006.01.004HilbertM.2016Big Data for Development: A Review of Promises and ChallengesDevelopment Policy Review341135174https://doi.org/10.1111/dpr.12142HoutkoopB. L.ChambersC.MacleodM.BishopD. V. M.NicholsT. E.WagenmakersE. J.2018Data Sharing in Psychology: A Survey on Barriers and PreconditionsAdvances in Methods and Practices in Psychological Science117085https://doi.org/10.1177/2515245917751886HuangR. L.WangX. K.HouW. H.ChenZ. Y.WangY. T.WangJ. Q.2023Two-way referral cooperative hospital selection with uncertain information: A two-sided matching decision-making approachComputers & Industrial Engineering183Article 109430https://doi.org/10.1016/j.cie.2023.109430KassN. E.NatowiczM. R.HullR. C.FadenR. R.PlantingaL.GostinL. O.SlutsmanJ.2003The use of medical records in research: What do patients want?Journal of Law Medicine & Ethics313429433https://doi.org/10.1111/j.1748-720X.2003.tb00105.xKowalczykS.ShankarK.2011Data Sharing in the SciencesAnnual review of information science and technology45247294https://doi.org/10.1002/aris.2011.1440450113LiX.LuJ. P.HuS.ChengK. K.De MaeseneerJ.MengQ. Y.HuS. S.2017The primary health-care system in ChinaLancet3901011225842594https://doi.org/10.1016/s0140- 6736(17)33109-4LiX. J.CongY. L.2024Exploring barriers and ethical challenges to medical data sharing: perspectives from Chinese researchersBmc Medical Ethics251Article 132https://doi.org/10.1186/s12910-024-01135-8PalmJ.SödergrenK.BockenN.2019The Role of Cities in the Sharing Economy: Exploring Modes of Governance in Urban Sharing PracticesEnergies1224Article 4737https://doi.org/10.3390/en12244737PrajogoD.ToyJ.BhattacharyaA.OkeA.ChengT. C. E.2018The relationships between information management, process management and operational performance: Internal and external contextsInternational Journal of Production Economics19995103https://doi.org/10.1016/j.ijpe.2018.02.019PriegoL. P.WarehamJ.RomasantaA. K. S.2022The puzzle of sharing scientific dataIndustry and Innovation292219250https://doi.org/10.1080/13662716.2022.2033178ReddyR. C.MishraD.GoyalD. P.RanaN. P.2024A conceptual framework of barriers to data science implementation: a practitioners’ guidelineBenchmarking-an International Journal311034593496https://doi.org/10.1108/bij-03-2023-0160ThoegersenJ. L.BorlundP.2022Researcher attitudes toward data sharing in public data repositories: a meta-evaluation of studies on researcher data sharingJournal of Documentation787117https://doi.org/10.1108/jd-01-2021-0015WeitzmanE. R.KaciL.MandlK. D.2010Sharing Medical Data for Health Research: The Early Personal Health Record ExperienceJournal of medical Internet research122Article e14https://doi.org/10.2196/jmir.1356XuQ. L.JiaoR. J.YangX.HelanderM.KhalidH. M.OpperudA.2009An analytical Kano model for customer need analysisDesign Studies30187110https://doi.org/10.1016/j.destud.2008.07.001ZinsC.2007Conceptual approaches for defining data, information, and knowledgeJournal of the American society for information science and technology584479493https://doi.org/10.1002/asi.20508